Linus’ Law: What greater collaboration could do for scientific research

February 25, 2021

Contributed Commentary by Aoi Senju

February 25, 2021 | Having spent the last decade with one foot in the door of the scientific research world, and the other foot in the door of the software world, I have viscerally experienced a fundamental flaw in how scientific research is done today. 

Collaboration among scientists is severely lacking.

When I start a software project today, I can look at my organization’s code repository to see if others on my team have already built something similar to what I’m trying to build. If my team hasn’t built anything similar, I can take a look at public code repositories to see if there might be any open-source projects that might point me in the right direction. 

I’m able to:

  1. clone these repositories
  2. alter them for my experiment
  3. call on my colleagues to help me validate the code 
  4. and merge the code onto my organization’s codebase, so that anyone at the organization can find and reproduce my work, even years later. 

The level of communication and collaboration throughout the software engineering process means that the code works as expected even after I’ve left the organization. 

This process exists at almost every software company in the world. 

But this does not exist in science! When scientists start new projects, they often start with reading through the existing literature, where they find a Methods section that’s often poorly written and/or unreproducible. Or, they dig through a colleague’s physical lab notebook (which might be gathering dust somewhere), and they try to decipher their colleague’s handwriting, try to find the page that the experiment was run on, or even discover that the experimental process had never been recorded. Oftentimes, the original scientist isn’t even there for consultation, because she or he graduated from the program (as all grad students do), or left the company years ago.

Information loss is obviously a problem within any organization, but it’s particularly an issue for the scientific community, with real financial consequences. There’s about $200 billion (about 85% of annual global spending on health and medical research) wasted on avoidable poorly designed and redundant studies. This comes as a result of a combination of factors, including that:

  1. half of all clinical trials in the US are never published in a journal (Nature, 2013)
  2. half of all published research have a poorly written Methods section (The Lancet, 2014)
  3. and half of all published research contain avoidable design flaws (British Medical Journal, 2015)

Scientists end up exploring well-trodden paths or wind up running up against dead ends that other scientists have already run up against, largely because of this lack of communication, lack of information sharing, and lack of oversight.

Imagine if every one of those unpublished papers had been available to the entire scientific community. Perhaps someone else could have come up with a different, more promising, conclusion. Imagine if every scientist had an “experiment review,” like the “code review” that software engineers have every couple of weeks? Perhaps these design flaws and unreproducible endpoints would have been caught. Imagine where science would be today? 

There’s a solution. Linus’ Law is an assertion in software development, that “given enough eyes, all bugs are shallow” — the more people who are shown the code, the more likely that the problems in the code will be uncovered by someone, and the more likely that the fixes to the problems will be obvious to someone. Scientific research just needs to embrace greater collaboration. 

There are examples of collaborative science happening today—even science that’s being shared online. We see this with the growth of online preprints by 142% in the last year, on sites like arXiv, MedRXiv, and BioRXiv, where unpublished research is regularly being posted to be reviewed and critiqued by the community, and with projects like PubPeer, which provides a forum for post-publication peer review.  

We’ve barely begun to explore all of the possible approaches to improving science, but it’s undeniable that greater collaboration has the potential to make science stronger. The reality is that today we live in an increasingly connected world where active communication is not only possible, but beneficial, for science. Collaboration widely exists outside of scientific research, and a simple implementation of a collaborative tool would radically transform science as we know it today. The COVID pandemic has already forced many labs to find online collaborative solutions. I suspect that even after COVID ends, many of these collaborative habits will remain in place.

Aoi Senju is Co-founder and CEO of Colabra, bringing the magic of collaboration to science (invested by Village Global, a venture firm backed by Bill Gates, Jeff Bezos, Reid Hoffman, etc.). He was trained as a chemical engineer and data scientist in the battery industry. He can be reached at aoi@colabra.app.